Course description

This course gives an introduction to probabilistic graphical models (PGMs), which is a probabilistic framework for modelling a system (or the joint distribution) over a large number of variables that interact with each other. The PGM framework lies at the intersection of statistics and computer science, combining concepts from probability theory, graph algorithms and machine learning. PGMs make up the basis of a wide variety of applications, including causal inference, medical diagnosis, and gene expression analysis. This course will focus on the two basic PGM representations: Bayesian Networks (directed graphical models) and Markov networks (undirected graphical models), and it will cover the three main topics related to PGMs: (i) Representation - the theoretical properties of a PGM, (ii) Inference - using a PGM to answer questions about the variables, and (iii) Learning - learning a PGM from data.